Beyond Polarity: Multi-Dimensional LLM Sentiment Signals for WTI Crude Oil Futures Return Prediction
Sophie Weber
Photo by Google DeepMind on Pexels
Swiss finance professionals may find relevance in the application of advanced language models to predict crude oil futures returns, a key commodity in glob
Beyond Polarity: Multi-Dimensional LLM Sentiment Signals for WTI Crude Oil Futures Return Prediction
A new paper by Dehao Dai proposes replacing the binary positive/negative sentiment scores commonly used in commodity trading with a five-dimensional sentiment framework extracted by large language models. Applied to WTI crude oil futures, the approach outperformed traditional sentiment indicators in predicting short-term returns across a dataset spanning 2020 to 2025.
Moving Past Simple Sentiment Polarity
Conventional sentiment analysis classifies news articles as positive, negative, or neutral. This reductionism discards most of the information in the text. A headline reading "OPEC agrees to production cuts amid weakening demand" contains both a supply-tightening signal (bullish) and a demand-deterioration signal (bearish). A single polarity score collapses these into one number, losing the tension between the two forces.
The multi-dimensional approach extracts five distinct sentiment channels from each article: supply-side sentiment, demand-side sentiment, geopolitical risk perception, macroeconomic outlook, and speculative positioning cues. Each dimension is scored independently, creating a richer input vector for downstream forecasting models.
The LLM Advantage in Nuanced Extraction
The researchers used large language models rather than traditional NLP tools to perform the extraction. The reasoning is straightforward: LLMs can parse complex sentences, understand implicit references, and distinguish between factual reporting and editorial speculation. When an article mentions "strategic petroleum reserve releases," an LLM can correctly classify this as a supply-side intervention rather than a generic positive or negative signal.
The paper benchmarks the LLM-extracted signals against both dictionary-based sentiment tools and BERT-class models. The multi-dimensional LLM signals produced statistically significant improvements in return prediction accuracy, particularly during periods of high market volatility where conventional sentiment measures tended to generate conflicting or flat readings.
Trading Strategy and Risk Management Applications
The five-dimensional signals lend themselves to more granular trading strategies. A portfolio manager could construct trades that are long when supply-side sentiment is bullish and demand-side sentiment is neutral, while hedging positions when geopolitical risk perception spikes independently of fundamental supply-demand dynamics.
For risk managers, the decomposed signals offer better early-warning capabilities. A simultaneous deterioration across three or more dimensions might trigger de-risking protocols earlier than a composite index that averages the signals together. The framework is not limited to crude oil. Any commodity with a rich news flow, including natural gas, industrial metals, and agricultural products, could be analyzed with the same multi-dimensional approach.
The study's five-year dataset captures multiple stress periods, including the pandemic recovery, the 2022 energy crisis, and post-2023 OPEC+ recalibrations, giving the results reasonable external validity for production deployment.
Disclaimer: This article is for informational purposes only and does not constitute financial advice. Consult a licensed financial advisor before making investment decisions.
Source
Original Article: Beyond Polarity: Multi-Dimensional LLM Sentiment Signals for WTI Crude Oil Futures Return Prediction
Published: March 12, 2026
Author: Dehao Dai
This article was automatically aggregated from ArXiv Computational Finance for informational purposes. Summary written by AI.
Related Articles
Disclaimer
This article is for informational purposes only and does not constitute financial, legal, or tax advice. SwissFinanceAI is not a licensed financial services provider. Always consult a qualified professional before making financial decisions.
This content was created with AI assistance. All cited sources have been verified. We comply with EU AI Act (Article 50) disclosure requirements.
AI Tools & Automation
Sophie Weber tests and evaluates AI tools for finance and accounting. She explains complex technologies clearly — from large language models to workflow automation — with direct relevance to Swiss SME daily operations.
AI editorial agent specialising in AI tools and automation for finance. Generated by the SwissFinanceAI editorial system.
Swiss AI & Finance — straight to your inbox
Weekly digest of the most important news for Swiss finance professionals. No spam.
By subscribing you agree to our Privacy Policy. Unsubscribe anytime.
References
- [1]NewsCredibility: 7/10ArXiv Computational Finance. "Beyond Polarity: Multi-Dimensional LLM Sentiment Signals for WTI Crude Oil Futures Return Prediction." March 12, 2026.
Transparency Notice: This article may contain AI-assisted content. All citations link to verified sources. We comply with EU AI Act (Article 50) and FTC guidelines for transparent AI disclosure.
Original Source
This article is based on Beyond Polarity: Multi-Dimensional LLM Sentiment Signals for WTI Crude Oil Futures Return Prediction (ArXiv Computational Finance)
